Last month a founder forwarded me a research deliverable. Three thousand words. Nine sections. Forty-seven tables. And forty-seven times, across forty-seven cells, the same three characters: N/A.
I read it twice. Once hunting for a thesis. Once hunting for a lie. There wasn't one. That was the strange part. The document was scrupulously honest โ every field it could not fill, it refused to fill. Under security assumptions: information insufficient. Under team stability: information insufficient. Under narrative durability: information insufficient. It had been produced by a two-stage pipeline: stage one extracts facts, stage two writes analysis. Stage one returned nothing. Stage two, being obedient, produced a flawless skeleton with no body inside it.
I have spent sixteen years in this industry. I have never seen a more faithful portrait of it.
There is a version of this story that reads as comedy โ the machine that dutifully formatted its own ignorance. I want to argue the opposite. That empty document is the most intellectually honest artifact to cross my desk this year, and the fact that it looks like failure while a thousand confidently-filled equivalents look like success is the central pathology of the current cycle.
The template economy
Research became an industry the moment it became a marketing asset. In 2017 you could count the serious analysts on two hands, they all published on Medium, and most of them were wrong. Today there are firms shipping twenty "deep dives" a day, each decked out in the same nine-part architecture: technical, tokenomics, market, ecosystem, regulatory, team, risk, narrative, transmission. The architecture is not the problem. The architecture is good. The problem is that architecture is cheap and evidence is expensive, and under deadline pressure every organisation optimises for the cheap thing.
So we get the ritual. A section header guarantees a section. A section guarantees paragraphs. Paragraphs guarantee the appearance of diligence. Nobody downstream checks whether the technical section contains a single verifiable claim about a byte of deployed code, because the shape of the document satisfies the shape of the expectation.
Here is the information-gain test I now apply to everything I read: remove the project's name and drop a competitor's in its place. If the report still reads true, it contained zero information about either project. Most reports fail this test. Not because they are wrong โ because they are unfalsifiable. They describe a category while pretending to describe an instance.
The blank pipeline failed a different test. It failed to produce anything. But it passed the hardest one: it never asserted something it could not support.
Technical: what an audit can and cannot see
When I audited ERC-20 implementations in Cape Town in 2017 โ three token projects, four months, nineteen-hour days โ I learned that the difference between a safe contract and a fatal one is often a single line of ordering. Two of those projects carried reentrancy exposure that would have drained them. I documented it publicly on GitHub, and I lost two clients and gained something more durable: the conviction that technical precision is a form of social protection. You do not audit code for the code. Trace the code back to the conscience behind it โ to the person who will lose rent money if you are lazy.
That is what a technical section is supposed to do. It should name the contract, the address, the function, the invariant, the failure mode. It should tell you what happens at the boundary โ when the oracle lags, when the sequencer stalls, when governance executes a parameter change at three in the morning on a Sunday. It should quote the diff.
A technical assessment with no address in it is not an assessment; it is a mood.
The blank document wrote information insufficient under technical. I trust that sentence more than a hundred pages of "robust architecture" with no commit hash attached.
Tokenomics: where honesty goes to die
Supply tables are the easiest thing in this industry to fabricate and the hardest to falsify, because the fabrication lives in the future tense. Team: eighteen percent. Early investors: twenty-two. Community: forty. Ecosystem fund: twenty. Vesting: "over four years." Everyone nods. Nobody asks the only questions that matter โ cliff or linear, on-chain or contractual, who holds the multisig, and what happens to the price when the first tranche unlocks into a book that is already thin?
I ran a workshop series through 2020 called DeFi for Everyone. Two hundred people, weekly, in a community hall with bad wifi. What I learned there reshaped how I write about incentives. Retail users do not lose money to complexity. They lose money to incentives they were told were yields. The APR on the screen was real. The dilution it implied was not disclosed. Between those two facts sat a room full of people who had a savings account, a phone, and nobody to explain the difference between a reward and a rebate.
So when a tokenomics section is blank, understand what is not being hidden. It is not that the data is unknown. It is that the data has not been verified. Those are different sentences, and only one of them is a scandal.
I will say the unpopular thing. The phrase "liquidity fragmentation" is not a technical description. It is a fundraising thesis dressed as one. Every cycle a dozen new venues appear, each with its own token, each promising to unify liquidity by splitting it again. The word describes the thing being created, not the problem being solved. When a pitch deck leads with fragmentation, ask which new product the fragmentation narrative exists to justify. The answer is usually sitting in the allocation table โ the same table the blank report was honest enough to leave empty.
Market and ecosystem: the decay nobody prices
There is a number I keep returning to. Binance Launchpad returns have compressed from roughly 100x at the peak to somewhere near 10x in recent cohorts. Read it as a price. Then read it again as a diagnosis. What decayed is not the asset; it is the traffic โ an exchange's ability to monetise a captive audience has been thinning for three years, and the launchpad multiple is the visible surface of that erosion. When the audience stops being captive, the entire distribution model built on top of it starts to crack, and no amount of narrative about ecosystem expansion repairs the arithmetic.
That matters for ecosystem analysis, because most ecosystem analysis is distribution analysis in disguise. Who depends on whom. Which protocol's TVL is a function of an incentive programme with an end date. Which "integration" is one governance vote away from being switched off. The dependency graph is the argument. I have seen documents with a hundred ecosystem logos and not one arrow drawn between them, because arrows require you to commit to a direction, and direction requires you to be wrong sometimes.
Regulatory: apparent clarity, real consolidation
MiCA arrived as a gift and landed as a filter. On paper, Europe now has the thing everyone claimed to want โ a rulebook. In practice, the stablecoin reserve requirements and the CASP compliance burden are a fixed cost, and fixed costs are regressive. A well-capitalised incumbent absorbs them as a line item. A five-person team building something genuinely useful in a small jurisdiction absorbs them as an existential question.
The clarity is real. So is the consolidation it quietly mandates. The rulebook never says "only large firms"; it just prices the entrance so that only large firms can afford to read it.
The blank template listed the Howey factors โ money invested, common enterprise, expectation of profit, efforts of others โ and marked every one information insufficient. That is, almost by accident, the correct answer for most tokens. The test is not answered by the token. It is answered by the conduct of the people who sold it. Without knowing who said what, to whom, in which channel, on which date, the analysis cannot exist. Anyone who produces a confident regulatory verdict without that record is not analysing law. They are guessing at it and charging for the guess.
Team, governance, risk: the sections that never get written honestly
Team sections are reputation laundry. Governance sections are participation theatre. Risk sections are the most interesting of the three, because a risk section is a confession and most organisations are institutionally incapable of confession.
Here is what a real risk section looks like. Not a matrix of colours. A sentence like: if the top three liquidity providers withdraw, the pool is insolvent in under six hours. Or: the oracle has four reporters and three of them share a hosting provider. Or: the upgrade key is held by two people who have not logged in since March. Those are sentences you can act on.
A risk written down in specifics is a risk you can mitigate. A risk coloured amber has been managed only in the sense that a pressed suit manages a wound.
I spent 2022 doing the opposite of research. Eighty percent of my own portfolio was gone, and I ran a group called Code & Conversation โ fifty one-on-one sessions with developers who could not sleep, plus a shared audit of the legacy code from projects that had failed. We read the graveyards. We kept finding the same thing: the bug was never exotic. A missing check. A bad assumption. A single-threaded dependency everyone had agreed not to look at.
The failure was always social before it was technical โ someone had raised the concern, and someone else had said later. That is what a risk section is for. It is the place where later gets written down as a date and a name.
Narrative and transmission: expectations are the real order book
The closing sections โ narrative, expectations, transmission โ are where analysis usually abandons rigour and starts writing poetry. Narrative durability. FOMO and FUD indices. Heat cycles. It is measurable, but not by vibes. It is measurable by the gap between what a project promised and what it shipped, tracked in public, over time.
The tradeable object in this market is not the token. It is the distance between expectation and delivery, and that distance is finite โ every narrative has a meter on it, and the meter runs down whether or not anyone is watching.
Transmission is the section people skip and shouldn't. When a protocol changes, the effect does not stay inside it. It lands on the venue that lists it, the bridge that routes to it, the stablecoin that settles it, the treasury that holds it. Draw the arrows. Then ask which arrow is load-bearing. Most failures are not the failure of a component; they are the failure of the assumption that the components were independent.
The contrarian turn
Here is where I turn on my own argument, because it deserves it.
The blank report is honest, yes. But honesty is not the same as usefulness, and there is a quiet arrogance in a framework that produces forty-seven N/A and calls the job done. A good analyst does not stop at information insufficient. A good analyst goes and gets the information. Files the request. Reads the commit log. Calls the person who deployed the contract. The blank template is a confession of a broken pipeline, not a moral achievement, and I should be careful not to romanticise paralysis just because it is sincere.
The harder question cuts at me too. Is my insistence on evidence itself a performance? There is a genre of crypto writing โ and I have written inside it โ that treats scepticism as a personality. The reflexive sceptic is as lazy as the reflexive bull, only better dressed. Both have outsourced the work to a prior.
So let me be precise about what I am actually claiming. I am not saying templates are evil. I am not saying the pipeline failed. I am saying a document that can tell the difference between verified and unknown is worth more than one that cannot โ and the industry has optimised almost everything except that distinction.
There is a version of decentralisation that has nothing to do with tokens. It is the decentralisation of verification โ the ability of a reader, anywhere, to check the claim in front of them against a primary source. We build bridges, not just blocks, between people, and a bridge that cannot be walked in both directions is a wall with better marketing.
This is not a technology problem. It is a provenance problem. Which is exactly why the work I have been doing since 2025 โ connecting decentralised identity to the verification of machine-generated content โ stopped feeling like a side project. Five thousand pilot users, two thousand blocked identity fraud attempts, and one realisation that reframes everything: in a world where text is free to generate, the scarce commodity is not the words. It is the signature.
We will need the same thing for research. Signed, timestamped, attributable analysis, where the reader can see not only the conclusion but the chain of custody that produced it โ which claims were verified, which were inferred, which were left blank, and why. Open source is not a license; it is a promise, and the promise only means something when the blanks are honest.
Education is the only true decentralized currency. It does not inflate. It cannot be seized. It compounds in the hands of whoever holds it. Every serious analyst I know came out of somebody else's patience โ a mentor, a thread, a badly formatted PDF written at two in the morning by someone who wanted them to understand.
So here is the question I am left holding, and I do not have an answer for it. When the next hundred-million-dollar project publishes its research suite, and every section is full, and every table is populated, and every verdict is confident โ will you ask which of those cells was verified, and which was merely filled?
Because the document that says nothing is at least telling you the truth. The dangerous one is the one that has learned to answer.